Pretest Preventive Medicine 9th (2001)

Biostatistics and Methods of Epidemiology Answers 41 and specificity measures the tests performance in people who do not have the disease. These definitions can be illustrated as follows: Disease Present Disease Absent Test positive True positive (TP) False-positive (FP) Test negative False-negative (FN) True negative (TN) Sensitivity = TP / (TP + FN) Specificity = TN/(TN + FP) Among people who have the disease, there are two possibilities: either the test correctly identifies them (TP), or it falsely classifies them as negative (FN). Thus, among those with disease, sensitivity measures how often the test gives the right answer. (A good way to remember sensitivity is by the initials PID: positive in disease.) Similarly, among people who do not have the disease, there are also two possibilities: either the test will correctly identify them as not having disease (TN), or it will falsely classify them as diseased (FP). Thus, specificity measures how often the test gives the right answer among those who do not have the disease. (A good way to remem- ber specificity is by the initials NIH: negative in health.) As opposed to sensitivity and specificity, which measure the tests per- formance in groups of patients who do and do not have the disease, pre- dictive value measures how often the test is right in patients grouped another way: by whether the test result is positive or negative. Thus, pre- dictive value of a positive test is the proportion of positive tests that are true positives [TP/(TP + FP)], and predictive value of a negative result is the proportion of negative test results that are true negatives [TN/(TN + FN)]. But predictive value is a little tricky because it also depends on the prevalence of the disease in the population tested. In this case, Dr. Stewells assembled groups of 100 patients with and without cholera, and the preva- lence is not given. Therefore, predictive value cannot be calculated in this question, and the correct answer is B, since specificity is 88%, not 12%. 18. The answer is c. (Greenberg, 2/e, pp 76-78. Rosner, 5/e, pp 58-60.) In this study of 1000 patients with profuse diarrhea, 343 of them had cholera. Thus, the prevalence of cholera (in this population) was 343/1000. (Note that this is not an incidence because we are measuring cases at a specific point in time,

42 Preventive Medicine and Public Health rather than new cases that occur over a period of time.) The predictive value of a positive result can thus be directly determined as TP/(TP + FP) = 312/(312 + 79) = 80%. Similarly, the predictive value of a negative result is TN / (TN + FN) = 578/(578 + 31) = 95%. Note here that this predictive value refers to the pre- dictive value of the test in patients admitted to the hospital with profuse diar- rhea. Since prevalence data from the general population are still lacking, the usefulness of this test in the general population is undefined. Predictably, the positive predictive value of this test in an asymptomatic population will be less. 19. The answer is b. (Greenberg, 2/e, p 83.) As the prevalence falls, more and more of those tested will not have cholera. This would change neither the sensitivity nor specificity of the test, which do not depend on disease prevalence, but would affect predictive value: as prevalence falls, predictive value of a positive result also falls, whereas predictive value of a negative result rises. This makes sense: as a disease becomes more and more un- likely, positive test results should be viewed with increasing skepticism, whereas negative results become increasingly believable. 20. The answer is a. (Greenberg, 2/e, pp 91-93.) Randomization is the use of a predetermined plan of allocation or assignment of subjects to treat- ment groups such that assignment occurs solely by chance. It is used to eliminate bias on the part of the investigator and the subject in the choice of treatment group. The goal of randomization is to allow chance to dis- tribute unknown sources of biologic variability equally to the treatment and control groups. However, because chance does determine assignment, significant differences between the groups may arise, especially if the num- ber of subjects is small. Therefore, whenever randomization is used, the comparability of the treatment groups should be assessed to determine whether or not balance was achieved 21. The answer is d. (Greenberg, 2/e, pp 121-126.) The study described was a case-control study. In this type of study, people who have a disease (cases) are compared with people whom they closely resemble except for the presence of the disease under study (controls). Cases and controls are then studied for the frequency of exposure to a suspected risk factor. In case-control studies, the validity of inferences about the causal relationship between the exposure (cigarette smoking) and the disease (lung cancer) depends on how comparable the cases and controls are for all variables that

Biostatistics and Methods of Epidemiology Answers 43 may be related to both the risk factor and disease under study (e.g., age, sex, race, place of residence, and occupation). Matching is a method to control for confounding in case-control studies to eliminate the effect of any extraneous variable that is not under study but may have an effect on the results. In clinical trials, or experimental/intervention studies, the investigators allocate the exposure. Correlation studies are used to compare disease frequencies between entire populations (as opposed to individu- als). For example, a correlation study could examine the consumption of animal fat and the rates of colon cancer among 20 different countries. 22-24. The answers are 22-b, 23-d, 24-a. (Greenberg, 2/e, pp 113-115.) The relative risk is defined as the incidence rate among the exposed (Ie) divided by the incidence rate among the nonexposed (Io). In this case, (120/100,000)/(10/100,000) = 12. The attributable risk (AR) is defined as Ie − Io = (120/100,000) − (10/100,000) = 110/100,000. The attributable risk percentage is equal to [(Ie − Io) / Ie] × 100 = (110/100,000)/(120/100,000) = 92%. If the prevalence of smoking was reduced to 10%, 11/100,000 excess cases of lung cancer due to smoking could be averted. We can calculate this by using the population attributable risk (PAR), which is defined by the attributable risk x prevalence of exposure in the population. If the preva- lence of smoking in the population is 20%, then the PAR is calculated as fol- lows: AR × prevalence of exposure = 110/100,00 × 20/100 = 22/100,000. If the prevalence of smoking is 10%, then AR × prevalence of exposure = 110/100,000 × 10/100 = 11/100,000. 25. The answer is c. (Greenberg, 2/e, pp 94-97. Hennekens, pp 206-208.) Intent-to-treat analysis, that is including in the final results all the subjects who were initially randomized to receive either the drug or the placebo, is the preferred method of analysis for intervention studies. Although it may be tempting to include only those who complied with the medication, the results can be misleading. This study is a classic example of this pittfall. Indeed, the study showed that the difference in mortality between those who did and did not adhere to placebo was even greater: 15 versus 28%. The difference persisted even after controlling for 40 different confounders. Thus, something related to compliance (with either the medication or the placebo) appeared to decrease mortality. Therefore, as a rule, remember that once randomization has been performed, all participants, regardless of their compliance, should be included in the results.

44 Preventive Medicine and Public Health 26. The answer is b. (Greenberg, 2/e, p 125.) Matching is a technique used in the design of the study to control for confounding. Subjects enrolled in a study are matched for age, gender, smoking, or any variable that is not being analyzed. This technique is not used for large cohort stud- ies as it would often be too time-consuming, restrictive, and expensive to find a match for each subject entering the study. Therefore, controlling for confounding is done in the analysis when a large group is recruited. Match- ing is mainly used when dealing with small case-control studies where the number of subjects enrolled would be too small to yield statistical results if stratified by subgroups. Randomization is used in clinical trials to control confounding (sample size needs to be large—see the answer to question 20). Matching cannot be used in correlation studies or cross-sectional stud- ies: these are descriptive studies to assess disease occurrence and they do not have control groups to test a hypothesis. 27-28. The answers are 27-b, 28-a. (Greenberg, 2/e, pp 94-97.) In order to maximize compliance, a researcher can assess the compliance of subjects by giving them either the active or inert medication for a certain period of time, before the randomization for the study has occurred. Noncompliant persons can be dropped from the entire study population and the compliant persons are then randominzed to receive either active or inert medication (placebo). This technique was used for the physicians health study to determine if the use of aspirin would reduce cardiovascular mortality. Keep- ing logs and frequent contacts from research staff can also help maintain compliance. The use of the placebo is to assess for responses that may sim- ply be attributed to receiving an intervention, whether active or inert. It has been shown that even patients who receive inert medication can do better than if receiving nothing. Therefore, they need to represent the control group of any clinical trial to account for the placebo ef fect. In a double- blinded study, both the investigators and the subjects are not aware of who is receiving active or inactive medication. This reduces the bias in the ascer- tainment of outcome. Intent-to-treat analysis refers to including all subjects who were initially randomized in the final analysis of results, compliers and noncompliers alike (see the answer to question 25). α and β error are used for statistical significance and do not affect the internal validity (i.e., are not a source of bias) of a study. Loss to follow-up, particularly if it is large or unequal between the intervention and control groups, can be a major source of bias for any prospective study, including clinical trials, if it is linked to the

Biostatistics and Methods of Epidemiology Answers 45 exposure, to the outcome, or both. As this study is likely to require a long follow-up period, every effort must be made to ensure complete follow-up. The incidence of lung cancer will not affect the internal validity of the study, but if it is low, it may affect the power of the study to measure differences between groups (because there may be an insufficient number of outcomes to reach statistical significance between the two groups). 29. The answer is e. (Greenberg, 2/e, pp 49-53.) Comparison of crude death rates of countries with different population compositions is fruitless. Adjusting both crude death rates to a standard population gives age- adjusted rates, which can be compared. Developed nations have higher crude death rates because larger proportions of their populations are elderly and thus have a higher probability of dying. Since rates account for population size, a larger population can be compared with a smaller one. Death rates are just one factor in evaluating health care systems. 30. The answer is b. (Hennekens, pp 20-25.) Because the association between the risk factor (use of smokeless tobacco) and the disease (oral lesions) is measured at a single point in time in a whole group of subjects, this is a cross-sectional study. A case-control study might be performed over a similar time period, but the sampling would be different: one sam- ple would be selected from among those baseball players found to have oral lesions (the cases) and a separate sample would be selected from among those players whose mouths were normal (the controls). In a cohort study, the habits of a group of players initially free of the disease would be measured, and these players would be followed over time to see how many develop the lesions. A clinical trial involves allocation of the subjects by the investigator (usually randomly) to one of two or more treatment groups. 31. The answer is d. (Greenberg, 2/e, p 18.) Cross-sectional studies allow one to estimate the prevalence (the number of existing cases at one point in time divided by the population at risk) but not incidence (number of new cases occurring over a period of time divided by the population at risk and the period of time at risk). The prevalence of mouth lesions is 80/110 (73%) in the users of smokeless tobacco. The relative risk, the incidence density, and cumulative incidence rates all apply only to cohort studies where the occurrence of disease in initially healthy subjects is examined

46 Preventive Medicine and Public Health over time among the exposed and the nonexposed. The odds ratio applies to case-control studies and cross-sectional studies. 32. The answer is c. (Hennekens, p 357.) The chi-square test can be used for statistical analysis of categorical data (no fractions are possible; number of persons are categorized as ill or not ill, dead or alive, etc.; and data are often presented in 2 × 2 tables). The odds ratio can be used as a measure of association. In this case, it is equal to (80 × 34)/(30 × 2). Because the asso- ciation between the risk factor (use of smokeless tobacco) and the disease (oral lesions) is measured at a single point in time in a whole group of sub- jects, no temporal association between the exposure and the outcome can be assessed. Furthermore, as this is not a cohort study in which subjects are chosen on the basis of exposure, there should be no expectations that the number of exposed persons would be similar to those who are not exposed. 33. The answer is c. (Hennekens, pp 200-201.) The importance of blind- ing, while it usually cannot be overemphasized, is not relevant in a study with total mortality as the end point: it is not possible to misclassify someone as alive when that person is really dead (except with fraudulent results). Power is not relevant in a study that shows a significant effect. If the results had failed to show a significant difference (p > 0.05) between the two groups, one may wonder whether the study had sufficient power. In a randomized study, the percentages of patients who actually had myocardial infarctions should be similar in the two groups. Total mortality is a much more impor- tant end point than mortality from coronary heart disease, but long-term follow-up is absolutely essential in determining whether a therapy is useful. Perhaps the new agent simply postpones mortality by a few days or weeks. 34. The answer is b. (Greenberg, 2/e, pp 64-68.) The secondary attack rate of a disease is the ratio of the number of cases of a specified disease among persons exposed to index cases divided by the total number so exposed. According to the data, 400 cases of pertussis occurred among 4000 fully immunized children who were exposed to a sibling who had the disease. The secondary attack rate, as a percentage, among fully immu- nized children after household exposure is, therefore, ᎏ400 × 100% = 10% 4000

Biostatistics and Methods of Epidemiology Answers 47 35. The answer is c. (Jekel, 1996, p 208.) The efficacy of vaccine, or the percentage reduction in the incidence of disease in vaccinated compared with unvaccinated subjects, is given by the expression Protection = (inci- dence in unvaccinated − incidence in vaccinated)/incidence in unvacci- nated = 100 ΄΂ ΃ ΂ ΃΅/ᎏ400 − ᎏ400 ᎏ400 × 100% = 75% 1000 4000 100 36. The answer is e. (Greenberg, 2/e, pp 98-99.) The relative risk is the ratio of the incidence rates of two groups who differ by some factor—in this instance, immunization status: ᎏIncidencᎏe rate amᎏong unimᎏmunized cᎏhildren Incidence rate among fully immunized children or ᎏ400 caseᎏs/1000 eᎏxposed cᎏhildren = ᎏ0.4 = 4 400 cases/4000 exposed children 0.1 37. The answer is d. (Ingelfinger, 3/e, p 58.) In decision analysis, utilities refer to the relative values placed on various outcomes that could be expe- rienced by the patients, not the physicians. For example, perfect health might be assigned a utility of 100, and death assigned one of 0. What, then, would the utility be for life with moderate back pain? With careful ques- tioning, one finds that most patients place a higher value on life with dis- ability than would be anticipated. Different techniques can be used to have persons quantify utilities for a given outcome. 38. The answer is d. (Hennekens, 2/e, pp 295-319.) All the choices listed are methods to control for confounding. Matching and restriction (exclud- ing smokers among cases and controls) can be achieved in the initial phase of designing the case-control study and before collecting information. Ran- domization is used for experimental studies. Once the data is collected, control for confounding can be performed in the analysis by stratification or multivariate analysis if there is a need to control for mutiple variables. In

48 Preventive Medicine and Public Health this example, we would first calculate the crude odds ratio from the 2 × 2 table including all cases and controls. We would then stratify the data by smoking status and calculate the odds ratio (OR) for each stratum (smok- ers and nonsmokers) as demonstrated in the following: All Subjects Smokers Nonsmokers Dis + Dis - Dis + Dis - Dis + Dis - Drinker a b ab a b d d Nondrinker c cd c Crude (unadjusted) OR ORs adjusted for smoking If the OR for the smokers and the nonsmokers is different than the crude OR, then confounding is present. In this situation, the adjusted OR would also be different than the crude OR. 39-41. The answers are 39-d, 40-c, 41-e. (Greenberg, 2/e, p 18.) The incidence of illness (as a percentage) is the total number of persons who have symptomatic illness divided by the total population at risk, and the calculation is (312 / 614) × 100% = 50.8%. The percentage of cases of Ger- man measles that were asymptomatic, or subclinical, is calculated by divid- ing the number of asymptomatic persons by the total number of infected persons. The calculation is [207 / (207 + 312)] × 100% = 39.9%. The information was stratified by age to determine if rates were similar. Age- specific infection rates were 90% in all age groups 0 to 59 years of age, while the rate was 10% in persons 60 years of age and over. The low attack rate in persons 60 and over suggests that this age group had developed immunity to German measles as a result of prior exposure at least 60 years before since there was uniform susceptibility in persons under 60. 4244. The answers are 42-a, 43-d, 44-d. (Greenberg, 2/e, pp 78-79. Hennekens, pp 331335.) There is a trade-off between sensitivity and speci- ficity of a test because there is overlapping of the normal population and the population with disease for most screening tests. The interval 0 µg/dL to 30 µg/dL includes all values of the normal population, but also some val- ues of the population with cancer; therefore, no value above 30 µg/dL will occur in individuals without disease. At this cut-off, the test will be 100% specific. However, you will miss some individuals with cancer, as some will

Biostatistics and Methods of Epidemiology Answers 49 have values between 20 µg/dL and 30 µg/dL. If the interval of 0 µg/dL to 25 µg/dL is chosen, then some persons with levels above 25 µg will have cancer and others will not. There will be some persons without cancer who will test positive (false-positives) and some persons with cancer will test negative (false-negatives). The last interval will be 100% sensitive as it will detect all cancers: there will be no false-negative tests. The trade-off is that it will be less specific: some persons without cancer will test positive (false- positive). Therefore, some confirmation of the test by another more specific method will be necessary before we can draw any conclusion. 45. The answer is e. (Pagano, pp 15-24. Rosner , 5/e, pp 28-39.) Line graphs are useful for presenting continuous data over time within different popula- tions. In fact, in most cases, the horizontal axis scale in line graphs represents time in year, months, and so forth. Frequency polygons are used to illustrate frequency distributions for discrete or continuous data. More than one set of data can be superimposed for comparison. The horizontal axis often repre- sents measure of the variable of interest (e.g., cholesterol) and the vertical axis represents the distribution either in numbers, relative frequency, or cumulative frequency. Histograms can also be used for this purpose, one set of data at a time. The horizontal axis should represent the true limits of inter- vals between data points (upper and lower limit) and the vertical axis should begin at zero. A bar chart is used to depict the frequency distribution of nom- inal or ordinal data. Only one set of data is represented for each chart. Pie charts can be used to illustrate relative frequencies of categorical data. (Note: all graphs represent hypothetical data.) Number of accidents 40 30 20 10 1950 1960 1970 1980 1990 Year A (Bar chart)

50 Preventive Medicine and Public Health Relative frequency 0.45 0.40 0.35 0.3 0.25 0.2 0.5 0.10 5 5.2 5.4 5.6 5.8 6.0 6.2 Height B (Histogram) 12% Asia Country of 70% U.S.A. 18% Europe origin for all tourists visiting city x C (Pie chart) 0.40 Women Men 0.35 Relative frequency 0.30 0.25 0.20 0.15 0.10 Height D (Frequency polygon)

Biostatistics and Methods of Epidemiology Answers 51 Number of cases × 1000 5 Disease B 4 Disease A Disease C 3 2 1 91 92 93 94 95 96 97 Year E (Line graph) 46. The answer is d. (Rosner, 5/e, pp 923.) These curves have the same mode, median, and mean (measures of location). However, the spread is different and can be assessed by computing the standard deviation (mea- sure of dispersion), which will be different for both curves. Although a large sample size tends to reduce variation and narrow a curve, it is not a summary numerical measure. 47. The answer is b. (Rosner, 5/e, pp 13-14.) This curve is skewed to the left (or negatively skewed). Such curves have median values that are larger than the arithmetic mean, and the mean also lies to the left of the median. This occurs when more outlying values are smaller than the mean, or points below the median tend to be further away from the median than points above. A curve is skewed to the right (or positively skewed) when the opposite occurs, and the mean lies to the right of the median. This curve only has one mode but is not symmetrical nor normally distributed. 48-50. The answers are 48-b, 49-d, 50-a. (Rosner, 5/e, pp 183, 239, 243-245.) Large sample sizes increase the precision of a study and decrease the width of the confidence intervals (CI). If the confidence interval includes one when assessing relative risks or odds ratio, it includes the null value. Therefore, the p value will be higher than 0.05 and the study will not reach statistical significance. The smaller the sample size, the larger the CI will be, and the more likely a study will be unable to (or have the power to) demonstrate a statistical difference between two groups, and will have a lower power . Also note that the smaller the dif ference between the null

52 Preventive Medicine and Public Health and alternative means, the larger the sample size will need to be in order to demonstrate a statistical difference and reject the null hypothesis. 51-53. The answers are 51-d, 52-b, 53-c. (Rosner, 5/e, pp 63-65.) This is an example of receiver-operator curves, or ROC curves. The horizontal axis (x) represents 1 − specificity, or the false-positive rate. This is plotted against the sensitivity on the vertical axis (y). There is always a trade-off between sensitivity and specificity as no test is ever 100% sensitive and 100% specific. Each curve can be used to determine the optimal cut-off point for the respective test. In general, the point closest to the upper-left corner, where sensitvity is highest and the false-positive rate is lowest, is chosen as the cut-off. The area under the curve is used to calculate the diag- nostic accuracy (best combined sensitivity and specificity) of the test, that is, the probability of correctly identifying disease or no disease based on the result of the test. The larger the area under the curve, the better the test. In this example, test C has the largest area under the curve compared to the other tests, and therefore would have the greatest diagnostic accuracy. 54. The answer is c. (Greenberg, 2/e, pp 161-163.) Sensitivity analysis is used in decision analysis to determine how much impact different proba- bilities of a particular event will have on the choice of choosing one inter- vention over another. Computer programs can compute and plot these data. The maximum quality-adjusted life expectancy or years (or QALYs) for surgery is 4.5 and for radiation is 4.2. QALYs are plotted for radiation therapy and surgery for different probabilities of mortality from surgery. As expected, mortality from surgery does not impact the QALYs obtained from radiation therapy. However, as mortality from surgery increases, the QALYs for that intervention decrease. If mortality did not impact QALYs for surgery, you would obtain a straight line with the y coordinate at 4.5. The threshold is the point at which both interventions intersect: decisions will be made above or below that point. In this case, surgery is superior to radi- ation if the mortality is below 11%. However, if the mortality from surgery is higher than 11%, then you gain more QALYs from radiation therapy. The sensitvity analysis from this example demonstrates that mortality rate from surgery is an important variable for determining the best strategy. 55. The answer is e. (Rosner, 5/e, pp 219, 243-244.) The null hypothesis (the odds ratio equals one) is not rejected. The confidence interval includes 1, and the p value is higher than 0.05. There is a 95% confidence interval,

Biostatistics and Methods of Epidemiology Answers 53 so the alpha was set at 0.05. If the alpha is set at 0.10, we want to be 90% confident that the interval limits cover the true value of the odds ratio. This would therefore narrow the width of the confidence interval. Conversely, if we were to choose an alpha at 0.01, or wanting to be 99% confident that the limits cover the true value, the confidence interval would be larger. 56. The answer is c. (Rosner, 5/e, pp 425-455, 466-487, 612-625.) Sim- ple linear regression examines the association between two continuous variables, the outcome/response variable y (the glomerular filtration rate), also called the dependent variable, and a predictor/explanatory variable x (the plasma creatinine), also called the independent variable. The line y = α + βx expresses the relationship and is called the regression line where alpha is the intercept and beta the slope. The ultimate objective is to pre- dict the value of an outcome based on the fixed value of an explanatory variable. In this example, we would be able to predict the glomerular fil- tration rate from a particular value of plasma creatinine, and thus deter- mine what is considered to be within normal limits. Multiple regression is used when we wish to examine the relationship between multiple depen- dant variables and the independent variable. The relationship is expressed as y = α + β1x1 + β2x2 + иии + ε. Logistic regression is used when y (the dependent variable) is not a continuous variable, but rather a dichotomous variable (for example, presence or absence of disease). The goal would then be to predict the presence or absence of disease based on a certain value of Colon cancer incidence (rate per 100,000) M K HI L D G B J AF CE Consumption of fat (grams) Correlation study of consumption of fat and colon cancer among persons in various countries (identified by letters)

54 Preventive Medicine and Public Health the predictor variable. Correlation analysis is used to determine whether there is a linear relationship between continuous variables that are treated symmetrically. It would not identify relationships that are nonlinear. It does not imply a cause and effect relationship, nor does it describe the nature of the relationship. It is used to analyze relationships in correlation studies of population. Is there a linear relationship between fat consumption and colon cancer, immunization rates and infant mortality? Each country rep- resents a point in the plot with a combination of outcomes x,y. 57. The answer is d. (Greenberg, 3/e, pp 112-115.) Both studies have the same relative risk (0.6/1.3 and 9.2/20 = 0.46). The intervention reduces the mortality by more than 50%. The relative risk reduction is expressed as risk of mortality in the intervention group (RI) − risk of mortality in the control group (RC)/RC. For study A: |0.6 − 1.3|/1.3 = 0.54 and for study B: |9.2 − 20|/20 = 0.54. It represents the proportional reduction in rates of a bad event between the inter vention group and the control group, and basically gives similar information as the relative risk. These two measures are useful to determine the magnitude of the effect of a given intervention. These measures can be misleading in assessing the clinical relevance of an intervetion because the overall impact of the intervention is highly depen- dent on the rate of mortality in the control. In study A, the rate in the con- trol group is very low. Thus, even if the relative risk is very high, the intervention is associated with little overall gain. The absolute risk reduc- tion (ARR) is expressed as RI − RC. It is the arithmetic difference between the two groups, or the same as attributable risk. We say reduction if the intervention reduces the risk and increase if the intervention increases a particular risk (not meaning a bad outcome). For study A, the ARR is equal to a reduction of 0.7%, and for study B, it is equal to a reduction of 10.8%. This measure gives a better picture of the impact of an intervention and how much benefit can be attributed to the intervention. We can see that the intervention used in study B would provide more benefit than the inter- vention used in study A. Numbers needed to treat (NNT) are expressed as 1/ARR. This gives us an estimate of how many patients will need the inter- vention before we can avoid one bad outcome, and can be useful for clini- cians to get a perspective on the intervention in their practices. 58. The answer is c. (Rosner, 5/e, pp 603605, 592-594.) Since the crude and the gender-adjusted relative risks are the same, you can conclude that

Biostatistics and Methods of Epidemiology Answers 55 gender is not a confounder (using the change-in-estimate definition of confounding). However, the relative risk for men is different than for women. We conclude that gender is an effect modifier. Effect modification is a different concept than confounding. Confounding is a nuisance factor that needs to be eliminated because it causes a distortion of the results, sim- ply because that factor is distributed unevenly in exposed and unexposed individuals. Effect modification provides important information: the magni- tude of the effect of a particular exposure on the outcome will vary accord- ing to the presence of a third factor, in this case, gender. It is not related to the fact that there may be more men than women in one group or another. A third factor can be both a confounder and an effect modifier if the adjusted risk differs from the crude, in addition to having different risks in women and in men. It may be neither a confounder nor an effect modifer if the adjusted and crude risks are the same and if the rates in men and women were the same. Finally, it could be only a confounder if the crude and adjusted risks differ, but the rates between men and women are the same. Stratification can be used to evaluate both confounding and effect modifica- tion: it will eliminate confounding and describe effect modification. 59. The answer is a. (Rosner, 5/e, pp 407-411.) The Kappa statistic is often used for reliability studies. For example, it can be used to assess inter- rater reliability, such as comparing the readings of mammography between different radiologists. It could also be used to assess intrarater reliability, such as comparing responses from participants on surveys given more than once over a period of time to evaluate reproducibility of responses. The chi-square will not give the degree of association and is used for categori- cal data. The student t test and correlation studies are used to analyze con- tinuous data. 60. The answer is b. (Rosner, 5/e, p 349.) The Wilcoxon rank sum test as well as the Wilcoxon signed rank test and the signed tests can be used when we cannot assume that the underlying population is of normal dis- tribution, especially when dealing with small samples. The signed test and the signed rank tests are the counterparts (for nonparametric distributions) of the paired t test, and the rank sum is analoguous to the t test for inde- pendent samples. A drawback of nonparametric methods is that they have less power than the methods used when normal distribution is assumed. The chi-square test is used for categorical data. The Mantel-Haenszel is a

56 Preventive Medicine and Public Health statistical method used to control for confounding. Analysis of variance is used to test the difference between the means of more than two indepen- dent samples. 61. The answer is b. (Greenberg, 2/e, pp 18-19. Hennekens and Buring, pp 6466.) Prevalence is equal to incidence times the duration of disease, or P = I × D. The longer the duration of the disease, the more likely it is to be present at any given time. If a disease has a high mortality rate (short dura- tion), it is unlikely to be counted at any time. Prevalence of disease will increase when a new treatment decreases mortality. A high incidence of dis- ease may or may not have an impact on prevalence: it will depend on its duration and mortality rate. 62. The answer is a. (Pagano, p 314.) The degrees of freedom for the chi- square distribution are calculated as follows: (rows − 1)(columns − 1). So, for a contingency table as the one illustrated 2 × 2, the degrees of freedom would be (2 − 1)(2 − 1) = 1. If we were using a 3 × 4 table, then the degrees of freedom would be (3 − 1)(4 − 1) = 6. 63. The answer is c. (Greenberg, 2/e, pp 222. Rosner , 5/e, pp 713-716.) This is an example of a Kaplan-Meier method, also called the product-limit method, of estimating survival. This technique takes into consideration that not all individuals may be followed until they experience the end point or failur e (in this example, death). Some may be lost to follow-up prior to failure (move away, refuse to continue to participte any longer, etc.), and others who have not experienced an end point may not have been followed for the whole observation period because they entered late in the course of the study. These are called censored observations (incomplete observation of a time to failure). Kaplan-Meier curves appear like uneven steps. Other methods can be used (actuarial method), but the Kaplan-Meier is the most frequent. 64-67. The answers are 64-d, 65-c, 66-b, 67-e. (Wallace, 14/e, p 49.) Fetal mortality is defined as the number of stillbirths per 1000 births of gestational age greater than 28 weeks. It evaluates fetal losses of the third trimester. Maternal mortality refers to the death of a woman from any cause related to or aggravated by pregnancy or its management. Direct maternal mortality relates to the death of a woman from obstetrical com- plications of pregnancy, labor, puerperium, from interventions, omissions,

Biostatistics and Methods of Epidemiology Answers 57 or treatment (such uterine rupture, prolapse, etc.). Indirect maternal mor- tality relates to conditions aggravated or caused by pregnancy, labor, or puerperium (diabetes, congenital heart disease, etc.) but not directly ob- stetrical. 68-69. The answers are 68-b, 69-c. (Greenberg et al., 2/e, pp 133-135.) Internal validity can be questioned if there is systematic (nonrandom) error in the way information is collected. Systematic errors include bias and con- founding. If a study suffers from lack of internal validity due to serious selection or information bias, and/or failure to control for confounding, the results should be questioned. External validity refers to whether the results (internally valid) of a study can be applied to the other populations. This is a question of judging whether the subjects in the study are similar to the population you are interested in applying the results to (such as patients in your clinical practice). Power refers to the capability of a study to detect statistically significant results. Reliability is synonymous with precision: even though the results in the study described in question 69 reached sta- tistical significance (the CI does not include 1), there is a very large confi- dence interval, suggesting that the study is not precise (increased random error). Lack of precision is often due to small sample sizes. 70-73. The answers are 70-b, 71-a, 72-d, 73-c. (Rosner, 5/e, pp 776779.) Use of the student t test to assess the difference between the mean systolic pressures of pregnant and nonpregnant women would be appropriate since the two groups are independent samples and the out- come variable is quantitative (continuous) and approximately normally distributed. In the study comparing the occurrence of hepatitis B surface antigen in medical and dental students, use of chi-square analysis would be appropri- ate because both the predictor and outcome variables are categorical and dichotomous; that is, students are classified by the presence or absence of the antigen and by medical or dental student status. The McNemar test is used for a matched pair of categorical data. To compare the levels of blood glucose in rats to whom a drug was administered, analysis of variance would be appropriate because six differ- ent groups are to be analyzed (two sexes and three drugs), where one vari- able is categorical (sex/drug) and the other is continuous (glucose level). Analysis of variance will permit evaluation of the effects and interaction of sex and drug on the glucose level.

58 Preventive Medicine and Public Health The paired t test is appropriate for comparing paired (e.g., before and after) measurements. Use of the regular (student two-sample) t test in this instance is inappropriate because the two samples are not independent— the same subjects are in each. 74-77. The answers are 74-b, 75-a, 76-e, 77-f. (Greenberg, 2/e, pp 80-81, 139-143.) Effect modification occurs when one factor modifies the effect on outcome of another. As an example, a high bilirubin seems to be a much stronger risk factor for bilirubin-induced brain damage if the baby is sick in other ways (see question 58.) Confounding occurs when the association between two variables is distorted by the fact that both are associated with a third. For example, the association between coffee and lung cancer is distorted by smoking: among nonsmokers and smokers considered separately, coffee and lung cancer may be completely unrelated, but when the two groups are combined, an association appears to be present. Similarly, lead levels need to be related to IQ separately at each level of socioeconomic status to assure that the asso- ciation is not due to confounding. The possibility that hyperactive children have high lead levels because they are hyperactive, rather than vice versa, is not confounding; it is simply a case in which the direction of causality is turned around (effect-cause). Nondifferential misclassification results in the mixing of two groups because the measure of either the exposure or the outcome was imprecise, for example, assessing precise diet habits by ques- tionnaires in a case-control study, and going back many years. Most people are unlikely to remember what and how much they ate years ago, and thus exposures will be similar in the cases and controls. Recall bias, a form of differential misclassification, is unlikely in this setting. Nondifferential mis- classification always biases results toward the null value. Lead-time bias refers to a distortion of the apparent efficacy of a screening program (see answer to question 16). 78-81. The answers are 78-a, 79-c, 80-b, 81-e. (Greenberg, 2/e, pp 15-19, 113.) The point prevalence is the proportion of people in a population who have a disease at a given point in time. The numerator is the number of existing cases of a disease; the denominator is the total population at risk of the disease at that point in time. In order to compare rates of disease or death in two or more groups that differ substantially in age, sex, or racial composition, adjustment or stan-

Biostatistics and Methods of Epidemiology Answers 59 dardization of the rates is necessary to remove the effects of those differences. The standardized mortality or morbidity ratio (SMR) is the ratio of the observed number to the expected number of deaths or cases of the disease. For exam- ple, age-specific rates of angina pectoris in nonsmokers can be applied to the age distribution of smokers to obtain the expected number of cases of angina pectoris in the smokers. The SMR of smokers for angina pectoris is the observed number of cases divided by the expected number so calculated. The cumulative incidence is the number of new cases of a disease that occur in a period of time divided by the population at risk during that time. The incidence density takes into consideration the length of time subjects participated in the study and the denominator is expressed in person-time of observation. The relative risk (or risk ratio) is the incidence of disease in subjects with a risk factor divided by the incidence in those in whom the factor is absent. (The denominator is not the incidence in the general population because then subjects with the risk factor would be included. If the risk fac- tor is uncommon and the relative risk is close to 1.0, the error involved in using the general population for the denominator is small. However, other risk factors, for instance, a relative with CHD, are quite prevalent.) The term relative risk can be confusing when the risk factor has to do with being a rel- ative of a patient; in this instance, risk ratio is a preferable synonym. 82-86. The answers are 82-c, 83-e, 84-d, 85-b, 86-a. (Greenberg, 2/e, pp 80-82. Rosner , 5/e, pp 212-216.) Although these terms are usually applied to epidemiological studies, they are also applicable to examples from every- day life. Lead-time bias commonly refers to the apparent increase in life expectancy seen in patients who have their disease diagnosed with a screen- ing test. The problem is that the screening test does not actually result in the patients living any longer than they would have other wise; the fact is sim- ply that these patients are detected with the disease earlier in the diseases course. The same would be true of a study that found that anatomy students lived at the same address for a longer period of time than fourth-year med- ical students, most of whom move to start internships. The study would not be wrong, but any conclusions that suggested that anatomy students are more stable than fourth-year clerks would be meaningless. Surveillance bias refers to overdetection of the disease of interest be- cause one of the groups goes to the doctor (or has a diagnostic test) more often than does another group. Similarly, you are more likely to find some-

60 Preventive Medicine and Public Health thing that is lost in June (when you may be moving) than in March, when you are presumably in the middle of the term. Recall bias classically refers to a situation in which persons with a dis- ease are more likely to remember an exposure (say, to a toxic chemical) than persons who are healthy. This is part of a human tendency to look for explanations for bad outcomes—like failing an examination. A type 1 error occurs when a result is found to be statistically signifi- cant by chance in a sample even though there is no effect in the population. In the case of answering the question correctly, the chance of a type 1 error is 20% because even if you did not know anything about this question, you would have a 1 in 5 chance of getting it correct. Power is the chance of finding an effect in your sample if it truly exists in the population. One problem with finding out that your friends have been out at the movies is that they may not tell you the truth (recall bias), or you may ask the wrong ones, such as those sitting next to you in the library (surveillance bias). So you can give yourself credit if you made one of those choices as well, assuming you understood what you were doing! 87-90. The answers are 87-b, 88-e, 89-a, 90-c. (Greenberg, pp 2223, 49-53.) The case fatality rate is a measure of the severity of the disease. It is a ratio of the number of deaths caused by a disease to the total number of cases of that disease and is usually expressed as a percentage. The crude mortality equals the total number of deaths from all causes during a year divided by the average population at risk during that year. It is usually expressed as the number of deaths per 1000 people. The secondary attack rate is a measure of the contagiousness of an infectious disease. The numerator is the number of cases of disease in contacts of the index case; the denominator is the number of contacts exposed to the index case dur- ing a specified period. Rates of disease are called morbidity rates. 91-94. The answers are 91-b, 92-a, 93-e, 94-b. (Greenberg, 2/e, pp 19-21, 125. Rosner , 5/e, pp 591-596.) Matching is a way of selecting subjects that are comparable with respect to specific variables. For example, in a case- control study, a control could be selected that is the same age and sex as the case. It is thus a sampling strategy to achieve comparability among groups. Stratification is an analysis strategy with the same purpose. Thus, after the study has been completed, the subjects can be stratified, that is, divided into separate, relatively homogeneous strata, and the comparison between

Biostatistics and Methods of Epidemiology Answers 61 groups can occur within each stratum. For example, survival could be compared separately in different age strata, as in question 91. This might be important if the subjects with high renin levels were also older than the subjects with low levels, since a difference in survival between the two groups might be due to age, rather than to differing renin levels. Age adjustment takes stratification by age one step further. After mor- tality (or another parameter) is calculated for specific age strata, it is com- bined in a weighted average to yield a single number. The weights used are the sizes of the different age strata in a standard population. Age adjust- ment is used more often for comparing mortality in populations with dif- fering age structures. Multivariate statistical analysis, like stratification, is an analysis tech- nique for achieving comparability among groups. It involves modeling the associations between variables in order to allow their different effects to be isolated from each other. (For example, in multiple regression, the rela- tionships between variables are modeled as a straight line.) Survival analysis is a technique by which persons followed for variable lengths of time are counted according to the length of time they were fol- lowed. For example, in the cohort study of renin levels mentioned previ- ously, instead of simply comparing the proportions surviving five years, the cumulative probability of survival could be plotted for the two groups, and the two curves compared. The Kaplan-Meier and life table analysis are two methods used for survival analysis. The first plots the percentage of per- sons alive after each year since a diagnosis. 95-98. The answers are 95-d, 96-a, 97-b, 98-d. (Hennekens, pp 58, 132, 170.) For proper comparison of the frequency of a disease in two groups, the rate of disease, not the number of cases, must be compared. The number of cases may reflect the age structure of the population served by the hospital. Age-specific attack rates that incorporate the number of cases in each age group, divided by the number of persons in each group, should be calculated. In order to determine that an association between two conditions such as diabetes and obesity exists, an investigator must show that obesity is sig- nificantly more common in persons who have diabetes than in persons who do not have diabetes. The controls are necessary in order to test the significance of the association and must resemble the cases as closely as possible in all ways except for the absence of the disease under study.

62 Preventive Medicine and Public Health Whenever considerable numbers of a cohort are lost to follow-up, doubts about the validity of the conclusions arise. Because death may be a major reason for loss to follow-up, the most conservative approach is to assume that everyone lost to follow-up has died. Unless, in this example, the death rate in the anxiety neurosis cohort was still no greater than that in the general population (after adding another 50 deaths for the 20% of the 250 patients lost to follow-up), the conclusions are suspect. The conclusion in question 98 is invalid because of the lack of denom- inators to calculate the rate of bacterial endocarditis in different age groups. In addition, the autopsy series merely gives an estimate of the proportion of deaths in different age groups, not the frequency of occurrence of endo- carditis with age. The autopsy series may also be invalid as a source of data from which to draw conclusions because of factors that determined whether an autopsy was performed. 99-102. The answers are 99-b, 100-c, 101-a, 102-e. (Greenberg, 2/e, pp 159-164.) When constructing a decision analysis tree, the first node is a decision node to reflect the choices you have to make to manage a specific medical problem. Branches from the chance nodes must reflect all possi- bilities. Terminal nodes reflect the outcomes or utilities assigned to the out- comes, such as death, survival, quality-adjusted life years, and so forth. The tree is folded back fr om right to left to get the expected utilities for each choice of action. Thus, the utilities are expressed in the same units as the outcomes (e.g., probability of survival, quality-adjusted life years). Here, the utility for surgery is equal to (0 × 0.05) + {[(4.4 × 0.05) + (4.8 × 0.95)] × 0.95} = 4.5. Therefore, surgery provides more QALYs than radia- tion therapy. If radiation therapy was never associated with proctitis, then you use the QALYs associated with the branch no pr octitis, (3.5 × 0.40) + (5.0 × 0.60) × 1.0 = 4.4 103-106. The answers are 103-a, 104-d, 105-e, 106-c. (Hennekens, pp 73-88.) If the probability of an event is p, the odds of the event are p / (1 − p). The odds ratio is the ratio of the odds of exposure to the risk factor given disease (a/c) to the odds of exposure to the risk factor given no dis- ease (b/d). To illustrate that the odds of exposure given disease are a/c, the probability of exposure given disease is 5a/(a + c). So (1 − p) = c/(a + c), and the odds are [a/(a + c)]/[c/(a + c)], and the (a + c) phrases cancel out to give a/c. The odds ratio, therefore, is (a/c) / (b/d), which equals ad/bc.

Biostatistics and Methods of Epidemiology Answers 63 Odds ratios are mainly used in case-control studies, from which relative risk cannot be calculated directly. When the disease is rare, the odds ratio closely approximates the relative risk. However, the study in the example is a cohort study, so relative risk can be calculated directly from the table. It is equal to the risk (incidence) of suicide in those who served in Vietnam divided by the risk in those who served elsewhere, or [a/(a + b)]/[c/(c + d)]. Excess risk is defined as the difference between the risk in those with the risk factor and those in whom it is absent. Whereas the relative risk and odds ratio are unitless (since any measurements of time in the denomina- tors cancel out), the excess risk must have an explicit or implied time period in the denominator. In this example, a/(a + b) − c/(c + d) represents the excess risk of suicide in Vietnam veterans over a five-year period; it is five times as big as the excess risk for a one-year period. Thus, if the yearly risk of suicide was 0.2% in Vietnam veterans and 0.1% in other veterans, the relative risk would be 2.0, and the excess risk (risk difference) 0.1% per year, or 0.5% over the five-year period. The overall incidence of suicide (per five years) in the study is simply the number of suicides (a + c) divided by the population at risk (a + b + c + d). (Note that a more precise way to measure the incidence, relative risk, and so on would be to use person-years at risk in the denominators, but this leads to greater computational and conceptual complexity.) 107-110. The answers are 107-c, 108-e, 109-a, 110-d. (Pagano, pp 469-472.) Simple random sampling is a process in which individuals are sampled independently, and each individual of the population has an equal probability of being selected. In cluster sampling, groups of people (e.g., families, school classes) are selected at random, and then everyone in those groups is sampled. A com- mon analytic mistake is to pretend that subjects obtained in a cluster sam- ple were obtained in a simple random sample. This can lead to incorrect results because the subjects are not sampled independently. Systematic sampling is a process that first requires the arrangement of the group to be sampled in some kind of order. Then individuals are selected systematically throughout the series on the basis of a predeter- mined sampling fraction or constant determinant, for example, every fifth, tenth, or hundredth person in the ordered group. Although systematic sampling may seem almost the same as simple random sampling, it is much less desirable. For example, sampling every other subject from a list

64 Preventive Medicine and Public Health in which husbands and wives names appear next to each other (e.g., an alphabetical list) will bias the sample—if husbands were always first, the sample might include no wives and would rarely include both persons in a married couple. In stratified sampling, a population is divided into subgroups based on defined characteristics such as age, sex, or severity of illness, or any combi- nation of these; then random samples are selected from each subgroup. For example, you could take a random sample from a group of 15- to 19-year- olds, from a group of 20- to 24-year-olds and from a group of 25- to 29- year-olds from a total population of 14- to 29-year-olds. This is used particularly in situations where the distribution of each subgroup is not uni- form in the group as a whole (for instance, there may be only a few 14- to 15-year-olds, and they may be missed if you were to use a simple random sample of the 14- to 29-year-old group). This method allows you to make sure that persons from each subgroup are represented in your sample. In paired sampling, or matching, selection of one or more controls for each case is based on age, sex, time, time sequence, geographic location, or some other defined relationship to the case (so it is not random). For exam- ple, selection could be based on the next patient admitted after each case, the sibling nearest in age to each case, or the person who lives closest geo- graphically to each case. 111-115. The answers are 111-d, 112-c, 113-a, 114-b, 115-e. (Greenberg, 2/e, pp 7681.) The easiest (and best) way to answer problems like this is to write out the appropriate 2 × 2 table. Since we were not told how many women were tested, we can just make up a number—say, 200. We are told that half have a positive test: Test positive Infection No Infection 100 Test negative 100 ? ? 200 ? ? The next task is to fill in the remaining boxes. We are told that 45% of those with a positive test are infected with chlamydia:

Biostatistics and Methods of Epidemiology Answers 65 Test positive Infection No Infection 100 Test negative 100 45 55 200 ? ? and that 95% of those with a negative test are free of the disease: Test positive Infection No Infection 100 Test negative 100 45 55 200 5 95 150 50 This now allows us to say that for 200 women in the community, the fol- lowing 2 × 2 table would be correct: Test positive Infection No Infection 100 Test negative 100 45 55 200 5 95 150 50 We can now determine the tests operating characteristics (sensitivity and specificity) and the other parameters. Sensitivity is simply the proportion of women with chlamydia who will have a positive test (remember: PID = positive in disease), or 45/50 = 90%. Specificity is the proportion of women without chlamydia who will have a negative test (remember: NIH = negative in health), or 95/150 = 63%. The prevalence of the disease is simply the pro- portion of women with chlamydia, or 50 out of 200 = 25%. The predictive value of a positive test is the proportion of women with a positive test who have chlamydia; we were already told that this was 45%. Likewise, the pre- dictive value of a negative test is the proportion of women with a negative test who do not have chlamydia; we were already told that this was 95%. 116-119. The answers are 116-c, 117-a, 118-b, 119-d. (Greenberg, 2/e, pp 123-125, 175. Rosner , 5/e, pp 213-215. Pagano, pp 218-222.) A type 2

66 Preventive Medicine and Public Health error occurs when a study fails to reject the null hypothesis (of no effect) when it is in fact false. Any time a study fails to achieve statistical signifi- cance, a crucial question to ask is whether the study had enough subjects. Although 500 subjects per group followed for five years seems like a large number, only a tiny minority (perhaps 10 per group) would be expected to have a myocardial infarction. Thus, the sample size in this instance may have been inadequate to detect a meaningful difference between the groups. The ecologic fallacy occurs when associations among groups of sub- jects are mistakenly assumed to hold for individuals. Thus, although among communities, high rates of condom use may be associated with higher fertility rates (perhaps because condom use acts as a marker for sex- ual activity in general), among those who use the condoms, the fertility rate could in fact be zero. A type 1 error occurs when, just by chance, a statistically significant difference between groups is found. Studies attempting to correlate multi- ple risk factors with multiple diseases (particularly when there is no good biologic reason to suspect an association) are especially prone to type 1 errors. Looking for associations separately in different subgroups com- pounds the problem. Selection bias occurs when the subjects selected for the study are somehow not representative of the population from which they come. One trouble with selecting spouses for controls is that ones spouse is much more likely to share ones smoking habits than a person from the general population. Thus, since patients with lung cancer will be mostly smokers, smokers will be overrepresented among the controls, and smoking will look like a weaker risk factor than it really is. 120-124. The answers are 120-a, 121-c, 122-e, 123-b, 124-d. (Pagano, pp 711.) The scale of measurement is an important determinant of the amount of information in a variable and the type of statistical analy- sis that can be used. Dichotomous variables (like sex) have only two possi- ble values. Some variables may be artificially dichotomized, with subsequent loss of information. For example, a patient either survives five years or not; thus survival to five years is an example of a dichotomous variable. The variable could be made more informative, however, if the actual number of months of survival was specified. Nominal variables have more than two possible values, but no intrin- sic ordering. Race is the classic example; medical specialties also have no

Biostatistics and Methods of Epidemiology Answers 67 intrinsic ordering. Nominal and ordinal are often confused. Just remember nominal for no or dering. Ordinal variables are intrinsically ordered, but not in a quantitative way that allows one to say that there is a natural numerical distance between possible values. Thus, one value cannot really be subtracted from another. Examples are qualitative judgments such as worse, same, better or never , sometimes, always. Remember , ordinal means or dered. Interval scales are ordered, but with real numerical units; they can be subtracted from each other. An example is dates of birth: they are intrinsi- cally ordered and subtracting them gives meaningful numbers, but there is no intrinsic zero to the scale, so that dividing them does not make sense— one date of birth cannot be twice as big as another. Ratio scales are measurements in relation to a clear zero point. Thus, measurements on ratio scales can be meaningfully divided by each other. For example, one baby may weigh twice as much as another or have twice as high a platelet count. Absolute temperature is measured on a ratio scale, whereas temperature in Fahrenheit or Celsius is measured on an interval scale. 125-127. The answers are 125-a, 126-b, 127-e. (USPS Task Force, 2/e, pp xliiixliv .) Answering the first two of these questions is easiest if the results of Dr. Bluess research are displayed in a 2 × 2 table: Positive Blues test Depressed Not Depressed Negative Blues test Total 80 60 20 340 100 400 The sensitivity of a test is defined as the proportion of persons with a disease who have a positive test (positive in disease = PID): in this case, 80 out of 100, or 80%. This is the same as the likelihood that a person with depression will have a positive Blues test. The specificity of a test is defined as the proportion of persons without a disease who have a negative test (negative in health = NIH): in this case, 340 out of 400, or 85%. The like- lihood that someone with a negative test will be depressed (posttest prob- ability of disease) is surely less than the overall prevalence of depression in the population (10% = pretest probability of disease).

68 Preventive Medicine and Public Health 128-130. The answers are 128-c, 129-i, 130-j. (Greenberg, 2/e, pp 76-78.) The probability of a positive test is the sum of all positives, true positives (TP) and false-positives (FP). Similarly, the probability of a nega- tive test is the sum of all negatives, true negatives (TN) and false-negatives (FN). When applied to a population, calculations can be done as follows: True positives (TP): Sensitivity × Prevalence of disease False-positives (FP): (1 − Specificity) × (1 − Prevalence) True negatives (TN): Specificity × (1 − Prevalence) False-negatives (FN): (1 − Sensitivity) × Prevalence The probability of infection given that the test is positive is the defini- tion of positive predictive value (PPV) and can be described as TP/(TP + FP). The probability of no infection given a positive test can be described as 1 − PPV. The probability of no infection given that the test is negative is the negative predictive value (NPV), expressed as TN/(TN + FN). The probability of infection given a negative test can be expressed as 1 − NPV. The tree can be completed as follows: DNA probe Test positive Infection (TP ϩ FP) (PPV) Test negative No infection (TN ϩ FN) (1 Ϫ PPV) Infection (1 Ϫ NPV) No infection (NPV)

EPIDEMIOLOGY AND PREVENTION OF COMMUNICABLE DISEASES Questions DIRECTIONS: Each item below contains a question or incomplete statement followed by suggested responses. Select the one best response to each question. 131. A 6-year-old child is brought 132. Which of the following con- to the emergency room by her par- ditions has been associated with a ents on a Friday night because they false-positive Fluorescent Trepone- are concerned about rabies. A bat mal Antibody Absorption (FTA- was present in the childs bedroom ABS) test? when they arrived at their country home that evening. It started flying a. Tuberculosis around the head of the girl when she b. Mononucleosis entered her room and it ruffled her c. Lyme disease hair. The parents heard her scream, d. Viral pneumonia ran up to her room, and shooed the e. HIV infection bat out the window. Upon examina- tion, there is no visible bite or scratch 133. One of your patients, a 30- marks. Which is the most appropri- year-old developer, tells you he is ate intervention at this time? planning a trip to the Dominican Republic the following month. He a. Reassure the parents that there is no will need to travel in rural areas. risk of rabies given the history and Which is the most appropriate in- examination tervention for malaria prophylaxis for this patient? b. Consult public health authorities to determine the epidemiology of a. No prophylaxis rabies in that area b. Chloroquine c. Mefloquine c. Administer rabies vaccine and rabies d. Doxycycline immunoglobulin (RIG) e. Primaquine d. Administer rabies immunoglobulin 69 (RIG) only e. Administer rabies vaccine only Terms of Use

70 Preventive Medicine and Public Health 134. A 20-month-old child pre- 136. Which of the following vac- sents to your office with a mild viral cines is CONTRAINDICATED dur- infection. The results of examination ing pregnancy? are normal except for a temperature of 37.2°C (99°F) and clear nasal dis- a. Hepatitis B vaccine charge. Review of her vaccination b. Varicella vaccine records reveals that she received only c. Influenza vaccine two doses of polio vaccine and d. Tetanus toxoid diphtheria-tetanus-pertussis (DTaP) e. Rabies vaccine vaccine, and that she did not receive the measles-mumps-rubella 137. A 32-year-old farmer pre- (MMR) vaccine. The mother is 20 sents to the emergency room with a weeks pregnant. Her brother is crushing injury of the index finger undergoing chemotherapy for leu- and thumb that occurred while kemia. Which of the following is he was working with machinery in the most appropriate intervention? his barn. Records show that he re- ceived three doses of Td in the past, a. Schedule a visit in two weeks for and that his last dose was given DTaP when he was 25 years old. In addi- tion to proper wound cleaning and b. Administer inactivated polio vac- management, which of the follow- cine (IPV) and DTaP ing is the most appropriate preven- tion intervention? c. Administer DTaP, oral polio vaccine (OPV), and MMR a. No additional prophylaxis b. Administration of tetanus toxoid d. Administer DTaP, IPV, and MMR c. Administration of tetanus immuno- e. Administer DTaP and OPV and globulin only schedule a visit in three months for d. Administration of tetanus toxoid MMR and immunoglobulin 135. Prevention of human brucel- e. Administration of tetanus and diph- losis depends primarily on theria toxoid a. Pasteurization of dairy products de- rived from goats, sheep, or cows b. Treatment of human cases c. Control of the insect vector d. Immunization of farmers and slaughterhouse workers e. Destruction of infected animals

Epidemiology and Prevention of Communicable Diseases 71 138. Epidemics of typhus fever 141. Professional organizations rec- have been associated with war and ommend that all pregnant women famine for several centuries. What be routinely counseled about HIV factor was most important in the infection and be encouraged to be control of such epidemics follow- tested. What is the most important ing the end of World War II? reason for early identification of HIV infection in pregnant women? a. Eradication of Anopheles mosqui- toes a. A cesarean section can be planned to reduce HIV transmission to the b. Improved sanitation practices newborn c. Improved methods for handling b. Breast feeding can be discouraged food supplies to reduce transmission to the new- d. Disinfestation by use of DDT born e. Mass therapy with antibiotics c. Early identification of a newborn at 139. Immunization of preschool risk of HIV infection will improve children with diphtheria toxoid survival results in d. Counseling on pregnancy options, a. Protection against the diphtheria such as termination, can be offered carrier state e. Antiretroviral therapy can be offered b. Lifelong immunity against diphthe- to reduce the chance of transmission ria of HIV to the newborn c. Detectable antitoxin or immuno- 142. A 35-year-old patient comes logic memory for about 10 years to your office in early April for a routine examination. In the course d. Frequent adverse reactions of the history, he tells you that he e. Protection against infection of the plans to go turkey hunting in Nan- tucket, Massachusetts, for one week respiratory tract by Corynebacte- in May. He is concerned about Lyme rium diphtheriae disease. Which is the most appro- priate intervention for preventing 140. What is the recommended Lyme disease? interval in months between the ad- ministration of whole blood transfu- a. Vaccination sion and the measles-mumps-rubella b. Avoidance of bushy areas (MMR) vaccine? c. Tick check at the end of each day d. Protective clothing and DEET a. 0 e. Antibiotic prophylaxis for one week b. 1 c. 3 d. 6 e. 10

72 Preventive Medicine and Public Health Items 143-144 145. In the course of investigating a 24-year-old HIV-infected male, An 18-year-old sexually active the HBsAg is positive. He is cur- college student presents with com- rently asymptomatic, his physical plaints of lower abdominal pain examination is essentially normal, and irregular bleeding for five days. and his CD4 cell count is 800. She has no fever. She uses oral con- Which of the following tests is most traceptives as method of birth con- helpful in determining whether the trol. Upon examination, the cervix patient is in the acute phase of viral is friable, there is cervical motion hepatitis? tenderness and adnexal tenderness. The pregnancy test is negative. a. ALT levels b. HBeAg 143. Which is the most likely etio- c. HBsAg logic agent responsible for these d. IgG anti-HBcAg findings? e. IgM anti-HBcAg a. Neisseria gonorrhoeae Items 146-148 b. Chlamydia trachomatis c. Treponema pallidum You are a newly employed phy- d. Herpes simplex virus type 2 sician at a community hospital and e. Mycoplasma hominis have been given the responsibility of overseeing the infection control 144. She tells you that she had a program. You plan to conduct a similar episode two years ago. What prospective surveillance of nosoco- is her risk of infertility following mial infections of patients, hire in- this second clinical episode of fection control personnel, and begin pelvic inflammatory disease? an educational program for hospital personnel. a. <1% b. 5% 146. Based on national data, you c. 10% expect that the incidence of noso- d. 20% comial infections in your facility e. 40% will be a. <1% b. 1 to 2% c. 3 to 5% d. 6 to 8% e. 9 to 10%

Epidemiology and Prevention of Communicable Diseases 73 147. You expect the most com- 150. Which patient is most likely mon site of infection to be to become a chronic carrier follow- ing an acute episode of hepatitis B? a. Urinary tract b. Surgical wounds a. A newborn c. Respiratory tract b. A 20-year-old female following d. Bloodstream e. Gastrointestinal tract vaginal sexual transmission c. A 50-year-old male following rectal 148. The intervention most likely to decrease the transmission of nos- sexual transmission with a partner ocomial infections in your institu- positive for HBeAg tion is d. A 30-year-old health care worker following a percutaneous injury a. Adding proper ventilation systems e. A 40-year-old HIV-infected male b. Disinfecting sheets and towels with a CD4 cell count of 200 c. Decreasing the use of indwelling Items 151-153 catheters d. Enforcing adherence to hand wash- A 2-year-old child is brought to the emergency room with severe ing prostration, a temperature of 40°C e. Eliminating common sources of in- (104°F), and a few petechial lesions around the ankles. She had mild fection upper respiratory symptoms until her condition started deteriorating a 149. In the United States, the larg- few hours before. A Gram stain on est proportion of tuberculosis cases the buffy coat of blood reveals gram- occurs among negative diplococci. Treatment is promptly initiated. a. HIV-infected persons b. Injecting drug users 151. The case fatality rate for this c. Homeless persons clinical manifestation of disease is d. Foreign-born persons e. Incarcerated persons a. Less than 5% b. 5 to 15% c. 20 to 30% d. 40 to 50% e. Greater than 50%

74 Preventive Medicine and Public Health 152. Compared with the general 155. Which sexual partners should population, the risk of developing be informed of the exposure and an infection among household con- referred for evaluation? tacts is a. Current sexual partners only a. The same b. Partners of within 30 days b. 10 to 20 times greater c. Partners of within 60 days c. 50 to 100 times greater d. Partners of within 90 days d. 200 to 400 times greater e. Partners of within 120 days e. 500 to 800 times greater 156. A 7-year-old girl is brought to 153. The child had been attending your office by her mother because of a day care center. In addition to rec- a rash that appeared three days ago. ommending close surveillance for Her temperature is 37.2°C (99°F) early signs of illness, which of the and her face has an intense rash with following is the most appropriate a slapped-cheek appearance. The management of day care contacts? most likely etiologic agent is a. No further action a. Adenovirus b. Vaccination of children only b. Rotavirus c. Vaccination of children and adults c. Parvovirus d. Antibiotic prophylaxis of children d. Coxsackievirus e. Echovirus only e. Antibiotic prophylaxis of adults 157. To which patient would the MMR be safe to administer? and children a. A 15-month-old HIV-infected child Items 154-155 with a CD4 cell count of 700 A 25-year-old man presents b. A 25-year-old pregnant woman with a single, indurated, painless ul- c. A 12-year-old asthmatic on 20 mg cer on the penis that appeared two days ago. His most recent unpro- of oral prednisone daily for the last tected sexual contact was 21 days 20 days before. An immediate rapid plasma d. An 18-year-old with leukemia in reagin (RPR) test is negative. remission whose chemotherapy was terminated 1 month ago 154. The most likely diagnosis is e. A 17-year-old with a life-threatening anaphylactic reaction to egg inges- a. Syphilis tion b. Herpes c. Chancroid d. Lymphogranuloma venereum e. Donovanosis (granuloma inguinale)

Epidemiology and Prevention of Communicable Diseases 75 Items 158-159 160. A 19-year-old college student presents to the university student On a Friday afternoon, a 30- health center complaining of severe year-old nurse is brought to coughing spells for the last four employee health for evaluation fol- days, following initial symptoms of lowing a needle-stick injury that coryza and malaise. She is afebrile. occurred at the AIDS clinic. The Her medical history is uneventful, source patient is known to be in- and immunizations are up to date. fected with HIV and has advanced She is a member of the basketball AIDS. team. During weekends, she baby- sits a 10-month-old and a 2-year- 158. Which of the following fac- old. In terms of management of tors carries the greatest risk for contacts, which etiological agent is transmission of HIV to the health the most important to include in care worker? the differential diagnosis? a. Depth of the injury a. Streptococcus pneumoniae b. Stage of illness of the source patient b. Mycoplasma pneumoniae c. Presence of visible blood on the c. Bordetella pertussis d. Influenza virus needle e. Legionella pneumophila d. Use of gloves during the procedure e. Entrance of the needle into a vein 161. Which of the following infec- tions is transmitted chiefly from or artery of the source patient person to person? 159. Which is the most appropri- a. California encephalitis ate course of action for this health b. St. Louis encephalitis care worker? c. West Nilelike viral encephalitis d. Meningococcal meningitis a. Reassure her of the low risk of in- e. Eastern Equine Encephalitis (EEE) fection and offer no prophylaxis for HIV infection b. Offer single-drug antiretroviral ther- apy c. Offer two-drug antiretroviral ther- apy d. Offer triple-drug antiviral therapy e. Draw an HIV antibody test and refer her to the infectious disease special- ist first thing Monday morning

76 Preventive Medicine and Public Health 162. Widespread use of the Haemo- 164. As an epidemiological inves- philus influenzae type b vaccine has tigation officer for the Centers for resulted in a dramatic decrease in Disease Control and Prevention, the number of cases of meningitis you are contacted by a local health due to this bacterium. Which agent department. They inform you that is now the leading cause of bacterial a large number of persons have meningitis in children in the United acquired mild symptoms of influ- States? enza despite being vaccinated for the appropriate strain being cul- a. Streptococcus pneumoniae tured. You find that the cultured b. Group B Streptococcus pyogenes strain is the same as that incor- porated into the trivalent vaccine (hemolyticus) administered throughout the world. c. Non-type-b Haemophilus influenzae You also note that the strain had a d. Escherichia coli K-1 high case fatality rate in previous e. Neisseria meningitidis epidemics in China, where most new strains are isolated and identi- 163. The medical evaluation of a fied for vaccine preparations. The 25-year-old intravenous drug user most likely explanation for the out- reveals elevated liver enzymes and break noted by the local health a positive anti-HBsAg. The most department is likely cause of the abnormal liver profile is hepatitis a. Vaccine failure b. Antigenic drift a. A c. Antigenic shift b. B d. Herd immunity c. C e. Incomplete immunity from previ- d. D e. E ous rhinovirus infections

Epidemiology and Prevention of Communicable Diseases 77 165. A 38-year-old HIV-infected woman presents for follow-up evalua- tion. She is on antiretroviral therapy. She has no complaints. Her physical examination is normal. Her PPD is reactive at 2 mm. The chest x-ray is nor- mal. She has no history of past TB or recent known contact with infectious TB. She lives at home alone. Her CD4 + T cell count is 180/µL. Her previ- ous count was 175/µL. Prophylaxis is most appropriate for which of the following infections? a. Mycobacterium avium complex (MAC) b. Cryptococcus neoformans c. Mycobacterium tuberculosis d. Toxoplasma gondii e. Pneumocystis carinii 166. During the investigation of an outbreak of food poisoning at a sum- mer camp, food histories were obtained from all campers as indicated in the following table. Which of the food items was probably responsible for the outbreak? Proportion of Campers Who Developed Illness (Percent) Campers Who Ate Campers Who Did Not Eat Food Specified Food Specified Food a. Hamburger 61 48 b. Potatoes 70 35 c. Ice cream 40 50 d. Chicken 73 10 e. Lemonade 20 45 167. In 1999, the majority of cumulative cases of AIDS in the United States occurred in which exposure category? a. Men who have sex with men b. Users of intravenous drugs c. Women who have sex with women d. Hemophiliacs e. Persons who engage in heterosexual contact

78 Preventive Medicine and Public Health 168. Which of the following com- Items 171-173 plications has been associated with the recall of rotavirus vaccine? As medical director of a divi- sion of epidemiology in a state a. Guillain-Barr syndr ome health department, you are asked to b. Hemolytic anemia develop a hepatitis C awareness c. Febrile seizures campaign. You develop a document d. Intussusception with answers to the most frequently e. Neutropenia asked questions (FAQ) by medical providers. You follow the 1999 CDC 169. The time interval between recommendations. entry of an infectious agent into a host and the onset of symptoms is 171. Which group should you rec- called ommend for routine screening? a. The communicable period a. Pregnant women b. The incubation period b. Emergency medical personnel c. The preinfectious period c. Health care workers d. The noncontagious period d. Persons who ever injected illegal e. The decubation period drugs 170. An 8-year-old child is brought e. Household contacts of HCV-positive to the emergency room with pro- fuse, bloody diarrhea. The symp- persons toms started about three days ago, but gradually worsened. He has no 172. Which test should you rec- fever. His platelet count is 40,000. ommend for screening? The most likely source of the enteric infection is a. EIA for anti-HCV b. Immunoblot assays a. Fish c. Qualitative HCV RNA b. Chicken d. Quantitative HCV RNA c. Milk e. ALT levels d. Eggs e. Beef

Epidemiology and Prevention of Communicable Diseases 79 173. What is the most appropriate 175. The most likely source of counseling message to offer to HCV- infection is positive pregnant women? a. A coworker a. Cesarian section should be per- b. Food formed c. His wife d. Water b. The probability of transmission to e. His children the newborn is 5% 176. A 10-year-old boy with sickle c. Breast feeding should be discour- cell disease presents with headache, aged anorexia, and fever. He complains of pain in the right tibia and local d. Infants should receive IgG at birth inflammation is noted. Osteo- e. Infants often do poorly in the first myelitis is diagnosed. The most likely etiologic agent is years of life a. Listeria Items 174-175 b. Salmonella c. Shigellosis You are a public health physi- d. Cryptosporidium cian working at a city health de- e. Campylobacter partment and receive a report of a case of hepatitis A virus (HAV) in- 177. HSV-2 seroprevalence has fection in a 32-year-old man who increased by over 30% over the past lives with his wife and one-year-old two decades in the United States, twins. He is a self-employed con- suggesting a continuing spread of tractor who often eats on the run. herpes. Which of the following His wife works part-time at a book- other epidemiologic findings have store and his children attend day been shown by recent studies? care. He has no history of travel, eating raw fish, or known contact a. Only 50% of persons with HSV-2 with other cases of HAV infection. antibodies have been diagnosed with herpes 174. The first step in investigating this case is to confirm the diagnosis b. HVS-2 seropositivity correlates of HAV with with viral shedding a. A report of the history and exami- c. Over 95% of genital infections are nation from the treating physician caused by HSV-2 b. Stool cultures d. Recurrence rates for HSV-2 are the c. Total anti-HAV antibodies same as for HSV-1 d. IgM anti-HAV e. HAV RNA e. Most transmissions occur during the symptomatic phase

80 Preventive Medicine and Public Health 178. A wildlife worker presents to 179. The child is most at risk for the emergency room because he which of the following complica- was bitten on the hand by a raccoon tions? while trying to capture the animal, which appeared ill. He states he a. Pneumonia received a primary course of rabies vaccination 112 years ago when he b. Reyes syndrome first started his job. The wound is immediately thoroughly cleaned by c. Encephalitis the ER staff. It is small because he was wearing gloves. Which is the d. Orchitis most appropriate intervention for rabies prevention? e. Thrombocytopenia a. No further prophylaxis is necessary 180. Which of the following is the because of the recent vaccination most appropriate management of contacts? b. Administer rabies immune globulin (RIG) only a. Observation only for all contacts b. Vaccine for the mother, sibling, and c. Administer RIG and one dose of vaccine susceptible classmates c. Immune globulin for the mother, d. Administer one dose of vaccine only sibling, and susceptible classmates d. Immune globulin for the mother e. Administer two doses of vaccine and vaccine for his sibling and sus- Items 179-180 ceptible classmates e. Immune globulin for the mother A 5-year-old child presents to and sibling, and vaccine for the the health department clinic with susceptible classmates fever, malaise, and a vesicular rash that started 24 hours prior. He goes to preschool. He has one sister aged 3 and his mother is 38 weeks preg- nant. Both are susceptible.

Epidemiology and Prevention of Communicable Diseases 81 181. Consider the clinical presen- 182. Under which conditions tation of the newborn in the follow- should chemoprophylaxis for influ- ing figure. enza be considered? a. All nursing home residents and unvaccinated staff during an influ- enza A outbreak b. All nursing home residents and unvaccinated staff during an influ- enza B outbreak c. Only unvaccinated nursing home residents and staff during an influ- enza A outbreak d. Only unvaccinated nursing home residents and staff during an influ- enza B outbreak e. All nursing home staff and residents during an influenza B outbreak (Reproduced, with permission, from 183. For which patient is pneumo- Holmes KK, Sparling PF, Mardh P, et al., coccal vaccine PPV23 not benefi- Sexually Transmitted Diseases, 3rd ed., cial? New York, McGraw-Hill, 1999.) a. A 15-month-old HIV-infected child This most likely represents congen- b. A 20-year-old about to undergo a ital splenectomy for ITP a. Rubella c. A 70-year-old healthy female b. Syphilis d. A 5-year-old with sickle cell disease c. Toxoplasmosis e. A 10-year-old with nephrotic syn- d. Cytomegalovirus (CMV) e. Varicella drome who received the vaccine 5 years ago

82 Preventive Medicine and Public Health 184. Consider the epidemic curve illustrated in the following figure. 65 54 4 33 333 3 2 22 21 1 111 1 1 10 00 000000 0 7/28 7/26 7/24 7/22 7/20 7/18 7/16 7/14 7/12 7/10 7/8 7/6 7/4 Date of onset (Source: Massachusetts Department of Public Health.) The curve most likely represents a a. Common-source outbreak epidemic curve b. Propagated-source outbreak epidemic curve c. Continual-source epidemic curve d. Person-to-person outbreak epidemic curve e. Point-source outbreak epidemic curve 185. A 10-month-old child is brought to your office by the mother because of vomiting and profuse diarrhea for the last 24 hours. He has a temperature of 100°F and has signs of dehydration. No other person in the household is ill. The most likely etiologic agent responsible for the clinical syndrome is a. Adenovirus b. Rotavirus c. Parvovirus d. Coxsackievirus e. Echovirus

Epidemiology and Prevention of Communicable Diseases 83 186. The medical evaluation of a 188. Four-drug therapy is recom- 32-year-old HIV-infected patient mended as an initial approach to reveals a tuberculin skin test reac- treatment for active TB in HIV- tion at 5 mm. His chest x-ray is infected persons normal. He is currently taking anti- retroviral therapy which includes a. Always protease inhibitors. He has not pre- b. When multidrug-resistant TB ex- viously received therapy for tuber- culosis in the past nor has he had ceeds 4% in the community any known contact with persons c. When the patient has had previous infected with tuberculosis. Which is the most appropriate interven- therapy for TB tion for this patient? d. When the patient has had a known a. No preventive therapy for tubercu- exposure to multi-drug resistant TB losis e. When the CD4 cell count is under b. Izoniazid for nine months 200 c. Rifampin for nine months d. Rifampin and pyrazinamide for two 189. You are contacted by a local physician who wishes to inform you months that she diagnosed and confirmed a e. Streptomycin for six months case of hepatitis A in one of her patients, a 5-year-old who attends a 187. HIV-infected persons are at preschool center. She is concerned highest risk of having an active TB about the staff and children attend- infection resistant to ing the school center. Which is the most appropriate management of a. Izoniazid susceptible contacts? b. Rifampin c. Streptomycin a. Immune globulin to all staff and d. Ethambutol children e. Pyrazinamide b. Vaccine to all staff and children c. Vaccine to staff and immune globu- lin to all children d. Immune globulin and vaccine to staff and all children e. Immune globulin only to classroom contacts

84 Preventive Medicine and Public Health 190. A 22-year-old woman pre- 191. Which persons in your city sents to the obstetrical clinic for her are at highest risk of developing second prenatal visit. She is 28 severe infection? weeks pregnant. The examination is normal. She reports multiple sexual a. The elderly partners, but denies drug use. The b. Newborns and young children RPR done at the first prenatal visit 8 c. Diabetics weeks ago was 1:64 with a positive d. HIV-infected persons TP-PA (treponemal test). The clinic e. Pregnant women was unable to reach her. She does not recall ever being treated for 192. What public health advisory syphilis in the past, nor does she measure would you announce to remember any symptoms. The rapid prevent ingestion of contaminated RPR card test done at this visit is water? positive. She is allergic to penicillin. The most appropriate intervention a. Drink bottled water only is to b. Use faucet filters capable of remov- a. Send for an RPR titer and trepone- ing particles of 2.0 microns mal test to determine appropriate c. Boil water for 1 minute treatment d. Disinfect with chlorination e. Freeze and use thawed water b. Treat with erythromycin c. Treat with doxycycline 193. As medical director of a health d. Admit for desensitization and treat maintenance organization, you are asked to update screening recom- with penicillin mendations for enrolled members. e. Treat with ceftriaxone You find that recommendations are lacking in the field of sexually Items 191192 transmitted diseases. You decide to develop evidence-based screening Following multiple reports of guidelines for Chlamydia trachomatis. cases of Cryptosporidium parvum di- Which of the following criteria is the agnosed by private physicians, as most important for developing rou- medical director of City XYZ Health tine screening recommendations? Department, you conduct an epi- demic investigation leading to the a. Number of sexual partners conclusion that the city drinking b. Use of barrier methods such as con- water supply is contaminated with C. parvum. doms c. Contact with an infected person d. Presence of symptoms e. Age

Epidemiology and Prevention of Communicable Diseases 85 194. A 20-year-old male presents with complaints of dysuria and urethral discharge for three days. He engaged in unprotected vaginal intercourse 8 days ago with a new female sexual partner. She has no complaints. Exami- nation reveals a yellow urethral discharge. The gram stain is as follows: (Reproduced, with permission, from Holmes KK, Sparling PF, Mardh P, et al., Sexually Transmitted Diseases, 3rd ed., New York, McGraw-Hill, 1999.) For which of the following organisms can a presumptive diagnosis be made: a. Chlamydia trachomatis b. Treponema pallidum c. Ureaplasma unrealyticum d. Neisseria gonorrhoeae e. Herpes simplex virus infection

86 Preventive Medicine and Public Health 195. A 30-year-old Canadian im- 197. A 20-year-old college student migrant farmer consults with symp- presents to your office because she toms of night sweats, low-grade noticed two bumps on her vulva 1 fever, cough, and fatigue. He does week ago. She does not complain not smoke. He has a history of about pain or discharge. She has asthma. The chest x-ray required been sexually active with the same for immigration was normal five partner for one year. He has no months ago. He received the BCG symptoms. She has noticed one vaccine as a child. The skin test for similar lesion on his penis. Upon tuberculosis is positive at 15 mm. examination, you notice two condy- The most likely diagnosis is lomata acuminata of 0.5 cm in size at the fourchette. Which of the fol- a. Influenza lowing counseling messages is the b. Brucellosis most appropriate? c. Aspergillosis d. Mycobacterium bovis a. Treatment of her sexual partner will e. Mycobacterium tuberculosis reduce the risk of recurrence of her vulvar lesions 196. A healthy 2-month-old infant is brought to the office for routine b. Treatment of her vulvar lesions will child care. The child has a normal reduce her risk of developing cervi- growth curve. She received the first cal cancer dose of hepatitis B vaccine at birth as well as a dose of HBIG because c. Condom use is very effective in the mother was HBsAg-positive. reducing transmission of this infec- Which of the following vaccine tion series should be administered at this time? d. Recurrence of lesions is more fre- quent in the first year after initial a. MMR, OPV, DTP, Hep B diagnosis b. IPV, Hib, DTP, Hep B c. Hep B, DTaP, Hib, IPV e. Pap smear screening should be per- d. DTaP, Hib, IPV formed every six months e. IPV, DTaP, Hep B

Epidemiology and Prevention of Communicable Diseases 87 DIRECTIONS: Each group of questions below consists of lettered options followed by numbered items. For each numbered item, select the appropriate lettered option(s). Each lettered option may be used once, more than once, or not at all. Choose exactly the number of options indi- cated following each item. Items 198-200 Items 201-204 Match each group of diseases Various terms and parameters with the most common described are used in epidemiological studies mode of transmission. of infectious diseases. Match each statement below with the most ap- a. Water- or foodborne transmission propriate descriptive term. b. Zoonoses c. Person-to-person direct contact a. Immunogenicity b. Pathogenicity transmission c. Infectivity d. Airborne transmission d. Virulence e. Arthropod-borne transmission e. Incubation f. Sexual transmission 198. Rabies, psittacosis, salmonel- 201. Neutralizing antibody devel- losis. (SELECT 1 DESCRIPTION) ops in 95% of people after an attack of measles. (SELECT 1 TERM) 199. Measles, tuberculosis, influ- enza. (SELECT 1 DESCRIPTION) 202. Febrile respiratory tract dis- ease develops in approximately 80% 200. Cyclospora, Campylobacter, Yer- of children infected with influenza. sinia. (SELECT 1 DESCRIPTION) (SELECT 1 TERM) 203. Death occurs in approxi- mately 20% of cases of pneumococ- cal meningitis. (SELECT 1 TERM) 204. Approximately 50% of house- hold contacts of a child who has a common cold become infected. (SE- LECT 1 TERM)

Rate per 100,00088 Preventive Medicine and Public Health Items 205-207 Consider the epidemiologic curves of sexually transmitted diseases reported to the CDC in the United States over the last 15 years. Match each curve with the appropriate infection. a. Chlamydia b. Gonorrhea c. Syphilis d. Herpes e. Trichomonas f. Human papillomavirus infection g. Hepatitis B 400 350 300 250 1 200 150 100 2 50 3 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 Year 205. Curve 1 206. Curve 2 207. Curve 3

Epidemiology and Prevention of Communicable Diseases 89 Items 208-210 Items 211-213 Choose the most likely infec- Match each of the statements tious agent for each description of below with the hepatitis virus with the events following consumption which it has been most closely of food. associated. a. Staphylococcal enterotoxin a. Hepatitis A virus (HAV) b. Clostridium botulinum toxin b. Hepatitis B virus (HBV) c. Enterotoxic Escherichia coli c. Hepatitis C virus (HCV) d. Clostridium perfringens d. Hepatitis D virus (HDV) e. Salmonella typhimurium e. Hepatitis E virus (HEV) f. Giardia lamblia g. Crytosporidium 211. Case fatality rate can be as h. Campylobacter high as 20% if acute infection occurs during the third trimester of 208. Within 4 h after attending a pregnancy. (SELECT 1 VIRUS) church supper, 25 persons report the abrupt onset of nausea, vomit- 212. Coinfection must exist for ing, and diarrhea. (SELECT 1 replication and infection to occur. AGENT) (SELECT 1 VIRUS) 209. One week after arriving in 213. Chronic disease develops in Africa, 16 students develop vomit- over 50% of persons following an ing, severe diarrhea, and abdomi- acute infection. (SELECT 1 nal cramps lasting 2 to 3 days. VIRUS) (SELECT 1 AGENT) Items 214-216 210. One-third of the persons who attended a school banquet develop Match each of the diseases be- abdominal cramps and watery diar- low with the appropriate epidemio- rhea 8 to 12 h later. These symp- logic term. toms end within 24 h. (SELECT 1 AGENT) a. Hyperendemic b. Epidemic c. Endemic d. Enzootic e. Pandemic f. Epizootic

90 Preventive Medicine and Public Health 214. Lyme disease in the 1990s. Items 220-224 215. Cholera among Rwandan ref- Select the reservoir for each of ugees in 1994. the diseases below. 216. Influenza in 1919. a. Cattle b. Humans Items 217-219 c. Rodents d. Ticks Match each infection below e. Mosquitoes with the intermediate host involved f. Cats in transmission. g. Soil h. Vegetation a. Snail b. Swine 220. Nocardiosis. (SELECT 1 c. Fish RESERVOIR) d. Crab e. Dog 221. Hantavirus. (SELECT 1 f. Cattle RESERVOIR) g. Deer h. Sheep 222. Brucellosis. (SELECT 1 RESERVOIR) 217. Paragonimiasis (lung fluke disease). (SELECT 1 HOST) 223. Enterobiasis. (SELECT 1 RESERVOIR) 218. Toxocariasis (visceral larva migrans). (SELECT 1 HOST) 224. Toxoplasmosis. (SELECT 1 RESERVOIR) 219. Cysticercosis. (SELECT 1 HOST) Items 225227 For each disease, choose the most effective or principal means of control. a. Rat control b. Sanitation c. Immunization d. Vector control e. Deer control